Introduction

MR imaging agents with excellent pH-sensing capability and imaging sensitivity could become very useful tools for monitoring tissue acidosis that presents in many disease states including rapidly growing tumors1,2,3. Such pH-sensitive agents must have pKa values near the pH of interest, ideally around ~6.7, display large pH-induced changes in chemical shift and have a suitable MR sensitivity for imaging pH in vivo. Several pH-sensitive probes reported previously rely on magnetic resonance spectroscopy (MRS) for readout of pH4. These include 1H-5,19F-6 and 31P-based agents7,8. Chelated Gd(III) and Eu(III) complexes have also been developed as MRI agents for pH mapping based on pH-dependent relaxivity and chemical exchange saturation transfer (CEST) imaging9,10, respectively. Despite some success using the Gd(III)-based agents to image pH in vivo, these agents require a separate measure of agent concentration for pH calibration. CEST agents offer the possibility of obtaining a direct readout of pH using ratiometric methods but the image sensitivity of such agents remains problematical. Recently, it has been reported that the sensitivity of NMR insensitive nuclei can be significantly improved, >10,000-fold or more, by use of dynamic nuclear polarization (DNP) methods11. In DNP, the extremely high thermal polarization of free radical electrons is transferred to the nuclear spins of NMR-active nuclei, such as 13C and 15N, by microwave irradiation of a frozen sample. The transfer generates a non-equilibrium “hyperpolarized” (HP) spin state with dramatically improved NMR sensitivity that returns to thermal equilibrium as a function of spin-lattice relaxation time, T1. DNP has played a key role in following real-time metabolism of HP 13C-labeled substrates such as [1-13C]pyruvate in tumors, heart and liver12,13,14,15,16. Hyperpolarized chemical probes that can in real-time monitor the pH, H2O2 and redox state have also been investigated17,18,19. For example, Gallagher et al recently demonstrated that HP 13C-enriched bicarbonate (HP H13CO3−) can be used to image pH of tumors in mice15,20. The same principle was later applied to measure intracellular pH (pHi) in isolated perfused hearts, based on intracellular generation of HP H13CO3−/13CO2 by oxidation of HP [1-13C]pyruvate21. It is important to point out that the pH estimated using HP bicarbonate likely reflects a combination of intracellular (pHi) and extracellular pH (pHe) because CO2 can freely diffuse through cell membranes20. The short T1 of H13CO3− is also a potential drawback for clinical translation of exogenous HP H13CO3−. In another example, Jindal et al reported HP 89Y-complexes as potential agents for chemical shift imaging (CSI) of pH by MRI22. Potential drawbacks for these complexes include the low γ of 89Y (low imaging sensitivity) and the relatively small pH-dependent chemical shifts (Δδ~10 ppm from pH 5 to 9) observed for such systems.

Herein, we demonstrate the potential utility of using HP 15N-pyridine derivatives as ultrasensitive pH imaging agents (Fig. 1). We first investigated the 15N NMR signal enhancement by DNP of the sp2-hybridized nitrogens in pyridine derivatives (i.e. pyridine, 2,6-lutidine, 2-picoline, nicotinamide and 2,4,6-collidine) and the 15N chemical shift pH dependence. Previously, it has been shown that the 15N chemical shifts of certain nitrogen functional groups were highly sensitive to protonation and a sharp chemical shift change near the pKa was observed23. Combining the strong pH-dependent chemical shift with the potentially substantial signal enhancement by DNP, HP 15N-enriched agents could be ideal candidates for chemical shift imaging of tissue pH. So far, hyperpolarization of 15N-labeled compounds by DNP has been demonstrated with significant 15N NMR signal enhancement11,24. It has also been shown that some 15N-labeled functional groups such as 15N-enriched choline (~4 min)24 and 15N-enriched nitro compounds (~100 s)25 have much longer T1 than carbonyl 13C centers.

Figure 1
figure 1

15N NMR spectra of HP-pyridine as the free base (green) and fully protonated (red) forms.

The large chemical shift difference demonstrates that pyridine derivatives may serve as ultra-sensitive pH probes. The two spectra were acquired separately from HP-pyridine samples dissolved in either base or acid.

Full size image

Results and Discussion

15N-NMR signal enhancement and spin-lattice relaxation time of pH probes

In this study, we aim to develop HP spectroscopic imaging agents that are highly sensitive to pH changes. Pyridine derivatives were chosen for evaluation because it is known that their 15N resonances have relatively long T1's and are highly pH-sensitive23. The pKa values of pyridine derivatives are optimal for tissue pH measurements and can also be altered considerably by adding substituents. Therefore, it is highly possible that pyridine derivatives having suitable pKa values for any specific pH measurement applications can be designed and synthesized. Despite the excellent sensing capability, the role of 15N NMR in molecular imaging is very limited mainly due to its poor sensitivity. To overcome such shortcoming, we investigated DNP hyperpolarization as a means to improve the 15N NMR sensitivity of pH-sensitive pyridine derivatives. Our results showed that these pyridine derivatives can be well polarized by DNP with high 15N signal enhancements. After dissolution into water, the 15N signal enhancements of 5437 ± 200, 8702 ± 700, 7877 ± 400, 10065 ± 600 and 3598 ± 700 were observed for 15N-pyridine, 15N-2,6-lutidine, 15N-2-picoline, 15N-2,4,6-collidine and 15N-nicotinamide, respectively. This level of enhancement is comparable to previously reported values for aromatic 15N compounds25. The observed higher polarization of methyl-substituted pyridine compounds was likely due to their greater tendency to form frozen glasses when mixed with the DMSO/Sulfolane glassing agent25. 15N NMR spectra of hyperpolarized and thermal samples of 15N-pyridine are compared in Fig. 2a. Here, the 15N NMR spectra of the HP sample was collected from a single 10-degree pulse while the thermally polarized spectra represent a total signal from 256 scans using 90-degree pulses. It is important to note that both the HP and thermally polarized spectra were acquired from the same sample. The big difference in signal intensity demonstrates the tremendous signal enhancement achieved by DNP hyperpolarization.

Figure 2
figure 2

(a) 15N NMR spectra of a HP (top) and thermal (bottom) sample of 15N-pyridine. (b) array of 15N NMR of HP 15N-pyridine after dissolution in water; (c) Representative T1 decay of HP 15N-pyridine in water.

Full size image

Fig. 2b shows an array of representative 15N NMR spectra of HP 15N-pyridine collected every 5 s using a 5-degree pulse after dissolution with water. By fitting such time-dependent 15N signal intensity decay curves (Fig. 2c), the T1 of HP 15N-pyridine (pH 8.4), 15N-2,6-lutidine (pH 9.5), 15N-2-picoline (pH 8.5) and 15N-nicotinamide (pH 5.9) were found to be 41 ± 3 s, 31 ± 2 s, 38 ± 2 s and 22 ± 0.3 s, respectively (Supplementary Fig. S1 online). Due to poor water solubility, dissolution of 2,4,6-collidine required the use of methanol. Here, the measured T1 was 36 ± 4 s. The presence of methyl substituents on the pyridine ring results in a larger number of protons in close proximity to the nitrogen which results in more efficient dipole-dipole relaxation of the pyridyl 15N. Nevertheless, these T1 values are comparable to many 13C carboxyl groups at this same field currently used for metabolic imaging. The hyperpolarized 15N signal of pyridinyl derivatives decayed more rapidly near physiological pH (Fig. 3a), likely reflecting an added relaxation pathway by the exchanging protons as these molecules become partially protonated26. As expected, the T1 of 15N-pyridine measured in plasma at 37°C was shorter (~11 s, Fig. 3b). While undesirable, this should not prevent the in vivo application of 15N-labeled pyridine derivatives as pH probes because protonation equilibria are established nearly instantaneously. In comparison, the 13C bicarbonate/carbon dioxide equilibrium system has been used successfully for imaging tissue pH in vivo even though the T1 of the 13C atoms in these molecules were only ~10 s20.

Figure 3
figure 3

(a) Representative T1 decays of HP 15N-pyridine and 15N-2,6-lutidine at physiological pH; (b) and HP 15N-pyridine in rat plasma. The inset shows 15N NMR spectrum of HP 15N-pyridine in plasma obtained by summing the 15N signal over 10 scans.

Full size image

pH sensitivity of HP 15N-pyridine derivatives

15N NMR has long been known to have a great sensing capability for different chemical environments such as acidity23 and metal ion concentrations27. Protonation or metal chelation of some nitrogen-containing compounds can lead to a significant change in 15N chemical shift, making them suitable MR sensors for pH or metal ions. For pyridine derivatives, the electronic properties of the sp2-hybridized, aromatic nitrogen center are greatly altered upon protonation leading to a significant change in 15N chemical shift. 15N NMR spectra of HP 15N-pyridine samples adjusted to different pH values are shown in Fig. 4a. All five 15N-pyridine derivatives shift upfield upon protonation (Fig. 4b). The 15N NMR linewidth of each probe was narrow in both the fully protonated or fully deprotonated forms but broaden somewhat at pH values near their respective pKa's. This broadening reflects intermediate rates of exchange of protons between the 15N atom and water26. The pKa values estimated by fitting the pH titration curves were 5.17 ± 0.07 for pyridine, 6.60 ± 0.02 for 2,6-lutidine, 6.02 ± 0.05 for 2-picoline, 7.65 ± 0.05 for 2,4,6-collidine and 4.14 ± 0.02 for nicotinamide. These pKa values agreed well with values previously reported28. The 15N chemical shift differences between the free base and fully protonated forms were all in the range, 88–94 ppm (Table 1). The chemical shifts of all five 15N enriched compounds changed linearly with pH at least over ±1 pH units of each ligand pKa. The data demonstrate that these 15N probes are quite sensitive to changes in pH (~60 ppm/pH unit) and that structures can be modified to fine-tune the probe to readout any desired pH value from 5 to 8.5.

Table 1 T1 relaxation time, signal enhancement, pKa and chemical shift changes of HP 15N-pyridine, 15N-2,6-lutidine, 15N-2-picoline 15N-2,4,6-collidine and 15N-nicotinamide. T1, enhancement and pKa are reported as mean values ± SD
Full size table
Figure 4
figure 4

15N chemical shift versus pH for HP 15N-agents; (a) 15N-NMR spectra of hyperpolarized 15N-agents versus pH; (b) 15N-NMR titration curves of hyperpolarized 15N-agents; (c) correlation of pH calculated from the Hendersen-Hasselbalch equation (displayed above) and pH electrode.

δobs is the 15N chemical shift observed from NMR spectra, δprotonated is the 15N chemical shift of fully protonated pyridine, δunprotonated is the 15N chemical shift of basic pyridine, n is the Hill coefficient.

Full size image

To test the accuracy of these pH probes, a series of samples containing HP 15N-pyridine or 15N-2,6-lutidine with variable amounts of HCl added to the dissolution solution were examined by 15N-NMR (Supplementary Fig. S2 online). The two agents were chosen because of their long T1 values. For each pH probe, the test was carried out over a pH range near the respective pKa values where the 15N chemical shifts are extremely sensitive to pH. The pH of each sample was estimated from the observed 15N chemical shifts (Fig. 4b) and also measured using a pH electrode29. Fig. 4c shows plots of pH as measured by 15N-NMR versus a pH electrode. The results show an excellent correlation between the two measurements for both probes (R2 > 0.99 for 2,6-lutidine and R2 > 0.95 for pyridine), demonstrating the high accuracy of pH measurements by 15N NMR of HP-pyridine derivatives.

Magnetic resonance spectroscopic imaging of HP 15N-Pyridine

Chemical shift images of phantoms containing HP 15N-enriched pyridine were collected to test the applicability of applying HP 15N-pyridine derivatives to distinguish pH differences. In this experiment, a solution of HP 15N-pyridine was simultaneously injected into two NMR tubes inserted within a large NMR tube (see Supplementary Fig. S3 online for the phantom setup). One of the small tubes contained a predetermined amount of sulfuric acid while the other contained sodium hydroxide. The imaging plane was positioned axially covering both tubes of HP-pyridine solutions (see1H image in Fig. 5). 15N CSI data demonstrate that the 15N signals of both basic and acidic HP-pyridine compartments were detectable by MRI. 15N NMR spectra of selected voxels show 15N resonances at 297.4 and 207.5 ppm reflecting the free base and protonated pyridine samples, respectively. CSI images of both basic and acidic pyridine displayed good localization of these two signals within the tubes (see CSI and merged image). The results show that magnetic resonance spectroscopic imaging (MRSI) of HP 15N-enriched pyridine was able to distinguish different pH environments in adjacent spatial locations in a phantom. It is worth noting that the imaging of HP 15N-pyridine was very rapid and a similar image of thermally polarized 15N-enriched pyridine would have required a much higher 15N spin concentration and a longer imaging time. These results strongly emphasize the sensitivity advantage of HP 15N for molecular imaging applications over thermally polarized imaging practices. To the best of our knowledge, these results demonstrate for the first time the feasibility of 15N spectroscopic imaging of pH-sensitive agents, strongly emphasizing the sensitivity advantage of HP 15N for molecular imaging applications.

Figure 5
figure 5

15N MR images of HP-pyridine in basic (top left) and protonated (top right) forms.

Localization of the 15N images to the1H reference image is shown in the merged image.1H MRI of a phantom showing CSI grids is shown in the middle right panel and the 15N NMR spectra of highlighted voxels are shown in the bottom panel.

Full size image

Potential in vivo imaging applications of these agents include, but not limited to, pH imaging or MR spectroscopic assessment of tissue acidosis that is present in many diseases. However, many considerations must be taken into account before HP pyridine derivatives can be translated for in vivo evaluations. First, the HP pyridine agents should preferably be present as a free base or be partially protonated at the physiological pH. The long T1 of the free base will allow for the HP signal to be retained much longer during the pre-injection quality controls of the agents. Additionally, the pH regions of the targeted tissues should not be very close to the pKa of the agents in order to avoid the rapid signal loss from fast water proton exchanges. Therefore, the ideal HP pyridine derivatives should have a pKa that is slightly below the pH of acidic tissues to be assessed in order to provide a relatively large chemical shift window and relatively long HP signal lifetime. The toxicity of these agents is also an important aspect that needs to be considered. Although the toxicity of pyridine is a concern, there are several naturally occurring pyridine derivatives that can be considered. Some of these compounds are present in the biological systems and play key roles in human physiology. For example, nicotinamide is a building block for nicotinamide adenine dinucleotide (NAD) and nicotinamide adenine dinucleotide phosphate (NADP) while nicotinic acid (vitamin B3) and pyridoxyl derivatives (vitamin B6) are essential human nutrients. Another derivative, picolinic acid, is a product of the kynurenine pathway and is a neuroprotective compound30. Chemical modifications of these molecules will be required for potential in vivo pH imaging applications to alter their pKa values toward the desirable range while preserving their biocompatibility. Future work will be focused on biocompatible pyridinyl compounds with favorable pKa values for pH imaging of tissue acidosis.

Conclusion

We have demonstrated the potential of using HP 15N-pyridine derivatives as pH-sensitive probes for MRI. These molecules display large changes in 15N chemical shift with pH and have very sharp chemical shift versus pH titration curves. The combination of hyperpolarization and pH-sensitive 15N agents offers new opportunities to develop highly sensitive imaging agents with great sensing capability for the characterization of important biomarkers such as tissue acidity. Future work will be focused on in vivo pH imaging of acidic tumors.

Methods

Acquisition of 15N-NMR spectra of hyperpolarized 15N-agents

Unless otherwise noted, all chemicals and solvents were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used as received. A mixture of pyridine (6.2 M), 2,6-lutidine (4.3 M), 2-picoline (5.0 M), nicotinamide (2.7 M) or 2,4,6-collidine (3.78 M) and BDPA radical (40 mM) dissolved in 50 μL of DMSO-sulfolane (1:1 v/v) was polarized in a HyperSense polarizer (3.35 T, Oxford Instruments Molecular Biotools, UK) according to the manufacturer's procedures. The polarization was carried out at ~1.05 K with 94.055 GHz microwave irradiation for 2 h. The dissolution liquid of HP agents in distilled water (4 mL) from the polarizer was rapidly mixed with a pre-determined volume of hydrochloric acid or sodium hydroxide in a 10-mm NMR tube to achieve the desired pH. 15N-NMR spectra were acquired at room temperature (~23 °C) on a 400-MHz spectrometer using a 5 degree flip angle with a repetition time (TR) of 5 s. All 15N peaks were externally referenced with 15N-nitrobenzene (372 ppm). 15N MRI were acquired on a 9.4 T Agilent vertical bore microimager (Agilent, USA). MR images and spectra were processed using ImageJ (NIH, USA) and ACD/SpecManager (ACD Labs, Canada). The T1, signal enhancement and pH titration experiments were investigated with the natural abundant 15N compounds. 15N-labeled pyridine was used in T1 measurement in plasma and CSI experiments.

T1 relaxation times and signal enhancements of HP pyridine derivatives

In a typical protocol, a dissolution liquid (4 mL) of HP-pyridine derivative was rapidly transferred into a 10-mm NMR tube. 15N NMR acquisition was initiated once the transfer was complete. An array of 15N spectra with one spectra recorded every 5 s was obtained (TR = 5 s, flip angle = 5 degree). 15N signal intensity of the HP nuclei was normalized to the signal intensity of the first time point (t = 0 s). By fitting the NMR signal intensity as a function of time, T1 values were calculated using Equation (1)31, where M0 is the original magnetization, TR is the repetition time and θ is the flip angle. Liquid-state 15N signal enhancements of HP pyridine derivatives were measured by comparing 15N NMR signal of HP versus thermally polarized samples. A single 5-degree pulse was acquired for the HP signal. For thermally polarized signal, a 20-mM solution of a pyridine compound was used. 256 scans of 15N NMR spectra were acquired using 90-degree pulses with a 500second delay (>5 times T1). Enhancement levels were calculated as ratios of the 15N signal from the two polarized states, taking into account for the different 15N concentrations and number of scans (1 vs 256). PBS buffer solution (pH = 7.4) was used as a dissolution solvent for T1 measurements at the physiological pH. The final concentration of HP 15N-pyridine or 2,6-lutidine was 2 mM. The plasma used in T1 of HP 15N-pyridine was obtained by centrifugation of whole rat blood to remove red blood cells. In this experiment, the dissolution liquid of HP 15N-pyridine (2 mL) was well mixed with rat plasma (2 mL) in a 10-mm NMR tube. The final concentration of HP 15N-pyridine was 13 mM. 15N-NMR spectrum was acquired at ~37°C.

pH titration curves

Titration curves were created by plotting 15N chemical shift versus pH as measured from a pH meter. For the determination of pH by HP 15N-NMR, HP-pyridine or HP-2,6-lutidine was mixed with an unknown amount of HCl or NaOH in an NMR tube before acquiring 15N-NMR spectra. pH of the solution was calculated from the apparent 15N chemical shift using the Henderson–Hasselbalch equation and the pKa value estimated by HP-NMR titration. The pH values obtained from HP 15N-NMR were plotted against the values measured by a pH electrode.

Chemical shift imaging (CSI) of HP 15N-pyridine

The phantom for this experiment was a 25-mm NMR tube encasing two 10-mm NMR tubes, with one pre-added sulfuric acid (200 μL, 7 M) and another had sodium hydroxide (200 μL, 5 M). The phantom was inserted into a 25-mm NMR tube containing DI water (~10 mL). The imaging plane was positioned axially across the phantom (Supplementary Fig. S3a online). 15N CSI was acquired after the transfer was completed with a 10-s delay to allow for complete mixing. pH values of the HP-pyridine solutions measured by pH electrode in the small tubes were 2 and 11. 15N CSI parameters: CSI2d sequence (Agilent VnmrJ 4 Imaging, Agilent, USA), field of view (FOV) = 40 × 40 mm2; TR = 200 ms; TE = 1.30 ms; flip angle = 20°; number of average (NA) = 1. A1H reference image was acquired at the same slice position using a GEMS sequence.1H imaging parameters: FOV = 40 × 40 mm2; TR = 200 ms; TE = 4 ms; flip angle = 20°; NA = 1. Matrix = 8 × 8; voxel size = 5 × 5 × 15 mm3. The CSI data were processed to 128 × 128 matrix.